Kuaishou

Kuaishou Technology is a Chinese short-video and live-streaming company headquartered in Haidian District, Beijing.¹ It grew out of “GIF Kuaishou”, an app for creating and sharing animated images that co-founder Cheng Yixiao launched in 2011,² and it has been listed on the Main Board of the Hong Kong Stock Exchange since Feb 5, 2021.¹ In the AI stack it is known for Kling AI (可靈AI), its family of video generation models. Kuaishou opened full beta testing of Kling in the second quarter of 2024, the same quarter it open-sourced its Kolors text-to-image model,³ and it dates Kling’s official launch to June 2024.⁴

Kling went through more than 20 iterations in its first year, including Kling 2.0 for global users in April 2025 and the Kling 2.1 series at the end of May 2025.⁴ The Kling 3.0 series, announced on Feb 6, 2026, raised the longest single clip to 15 seconds and added native audio, with speech in English, Chinese, Japanese, Korean and Spanish Vendor-stated;⁵ Kling Video 3.0’s specs are compared with other video models in How AI Video Generation Works: Clip Length, Resolution and Audio by Model. In the second quarter of 2026 Kuaishou added native 4K output to the 3.0 series, released Kling 3.0 Turbo, and launched Kling MCP and a Kling CLI so that AI agents can call the models.⁶ Kling 4.0, launched on Oct 15, 2026, generates up to 30 seconds in a single pass and accepts up to 15 reference assets and up to 10 keyframes, with output up to 10-bit HDR Vendor-stated; Kuaishou says API access for enterprises and developers will follow.⁷

Kuaishou reports Kling’s revenue separately. Kling’s annualized revenue run rate passed USD 100 million in March 2025, its tenth month after launch,⁴ and in the second quarter of 2026 Kling generated revenue of over RMB 850 million, up more than 200% year on year, as reported in Kuaishou’s interim results Vendor-stated.⁶

Kuaishou also publishes open models. LivePortrait, from researchers at Kuaishou Technology with the University of Science and Technology of China and Fudan University,⁹ animates a single portrait photo using the facial expressions and head pose of a driving video. Its paper, first posted on Jul 3, 2024, uses an implicit-keypoint framework rather than a diffusion model, scales training to about 69 million frames, and reports a generation speed of 12.8 ms on an RTX 4090 GPU with PyTorch Vendor-stated.⁸ The code and weights followed on Jul 4, 2024 under the MIT license; the repository, first published under KwaiVGI, now sits under KlingTeam on GitHub and Hugging Face, and its maintainers say major video platforms, including Kuaishou, Douyin, Jianying and WeChat Channels, have adopted it.⁸⁹¹⁰ On Hugging Face the LivePortrait weights were downloaded 15,798 times in the 30 days to Oct 18, 2026 Measured.¹⁰